Fair redistricting: Texas

[jimrtex]

First round - due to the large number of SC maps, each panelist was required to reject at least 5 of them. A map required at least two approving votes to pass. ASV-B, True Federalist-B, muon2-B, weatherboy-A and -B were eliminated. Here's the breakdown:

Sol
cvparty-A - N
cvparty-B - N
AustralianSwingVoter-A - Y
AustralianSwingVoter-B - N
Singletxguyforfun-A - Y
Singletxguyforfun-B - Y
MB - Y
weatherboy1102-A - N
weatherboy1102-B - N
True Federalist-A - Y
True Federalist-B - N
muon2-A - N
jimrtex-A - N
jimrtex-B - N
muon2-B - N

"I was a good bit harsher here, naying any map without a clear black-dominated district. This was my most important criterion. muon2-B was on the fence; with a BVAP 2 or 3 points higher I might have approved it."

me
cvparty-A - Y
cvparty-B - Y
AustralianSwingVoter-A - N
AustralianSwingVoter-B - Y
Singletxguyforfun-A - N
Singletxguyforfun-B - Y
MB - Y
weatherboy1102-A - N
weatherboy1102-B - Y
True Federalist-A - Y
True Federalist-B - N
muon2-A - Y
jimrtex-A - Y
jimrtex-B - Y
muon2-B - N

I believe I rejected the maps I did for splitting up metro areas weirdly, mainly Charleston. But feel free to ask about the rationale for my vote on a particular map.

Singletxguyforfun
cvparty-A N
cvparty-B Y
AustralianSwingVoter-A Y
AustralianSwingVoter-B N
Singletxguyforfun-A Y
Singletxguyforfun-B Y
MB N
weatherboy1102-A Y
weatherboy1102-B N
True Federalist-A Y
True Federalist-B N
muon2-A N
jimrtex-A N
jimrtex-B N
muon2-B N

no comment.

TimTurner
Y: cvparty-A
Y: cvparty-B
AustralianSwingVoter-A
AustralianSwingVoter-B
Y: Singletxguyforfun-A
Singletxguyforfun-B
Y: MB
weatherboy1102-A
weatherboy1102-B
Y: True Federalist-A
Y: True Federalist-B
Y: muon2-A*
Y: jimrtex-A
Y: jimrtex-B
Y: muon2-B

no comment.

I think these are the ten finalists.

cvparty A



cvparty B



AustralianSwingVoter A



Singletxguyforfun A



Singletxguyforfun B



MB



True Federalist A



muon2 A



jimrtex A



jimrtex B

[jimrtex]

First round - due to the large number of SC maps, each panelist was required to reject at least 5 of them. A map required at least two approving votes to pass. ASV-B, True Federalist-B, muon2-B, weatherboy-A and -B were eliminated. Here's the breakdown:

I suggest that next time, that panelists be limited to approving no more than half (i.e. half of 15 is 7.5 is and that you get a fifth panelist.

As it was, there were 31 approval votes (10, 10, 6, and 5), which appropriately distributed could have result in all 15 plans advancing. If all 4 panelist had approved 10, then you have the possibility of 40 approvals of 2.67 per plan.

Requiring 7 or 8 approvals would make the more discriminating panelists consider which of the other plans were the least worst; and the more generous panelists consider which of their approved plans were the least best.

With 5 panelists approving 8 plans, you would have 40 approvals or 2.67 per plan, but if a plan required 3 approvals it would cut down on the advancers.

Would it be possible to get a group of Democratic panelists and let them rank only one State at a  time?

Torie
Kevinsat
etc.?

If you only get three instead of four, then only do three states at a time.

[muon2]

I'll try to get the SPICE scores for the finalists here:

Plan	-S-	-P-	-I-	-C-	-E-
cvparty A	1	11	5	15	38
cvparty B	1	11	5	13	36
Aussie A	3	13	4	12	42
Single A	3	13	5	15	45
Single B	3	13	5	13	35
MB	2	11	10	12	39
True A	3	13	3	16	39
muon2 A	2	11	11	5	34
jimrtex A	4	10	5	7	39
jimrtex B	4	10	12	3	38

nb. The SPICE rules reward plans that when they chop as little as possible into a county. This proves to be very valuable in urban areas that have dense populations: the chop may look small in area, but has a large impact in population. A macrochop is a chop that includes more than 5% of a district's quota. When a county is macrochopped in SC I will treat the school districts in that county as the subunits, since there are no townships as there are in the northeastern US. When a macrochop occurs, the school districts in that county act is if they were the counties in terms of chops and erosity.

nb2. When a small chop goes into a county it is best to place it along the road that connects the county seats, or at least along a state or federal highway. The idea is that the fragments of a chopped county should be well connected to the other counties in the same district.

[GM Team Member and Senator WB]

wait... according to your post, tim didn't vote on my two maps.  How does that work?

[cvparty]

wait... according to your post, tim didn't vote on my two maps.  How does that work?

He did, he just put Y for the 10 (maximum) maps he could approve. The remaining 5 are presumably the NO votes

[cvparty]

First round - due to the large number of SC maps, each panelist was required to reject at least 5 of them. A map required at least two approving votes to pass. ASV-B, True Federalist-B, muon2-B, weatherboy-A and -B were eliminated. Here's the breakdown:

I suggest that next time, that panelists be limited to approving no more than half (i.e. half of 15 is 7.5 is and that you get a fifth panelist.

As it was, there were 31 approval votes (10, 10, 6, and 5), which appropriately distributed could have result in all 15 plans advancing. If all 4 panelist had approved 10, then you have the possibility of 40 approvals of 2.67 per plan.

Requiring 7 or 8 approvals would make the more discriminating panelists consider which of the other plans were the least worst; and the more generous panelists consider which of their approved plans were the least best.

With 5 panelists approving 8 plans, you would have 40 approvals or 2.67 per plan, but if a plan required 3 approvals it would cut down on the advancers.

Would it be possible to get a group of Democratic panelists and let them rank only one State at a  time?

Torie
Kevinsat
etc.?

If you only get three instead of four, then only do three states at a time.

I was wary of rejecting too many maps and figured too much would be better than too little. But yes I'll probably try to narrow it down further next time. And I am trying to find a good fifth panelist

[GM Team Member and Senator WB]

georgia will be a mess if I remember their voting blocs in cobb county correctly.

[True Federalist (진정한 연방 주의자)]

I'll try to get the SPICE scores for the finalists here:

Plan	-S-	-P-	-I-	-C-	-E-
cvparty A
cvparty B
Aussie A
Single A
Single B
MB	2	11	10	12	39
True A	3	13	3	16	39
muon2 A	2	11	11	5	34
jimrtex A
jimrtex B

nb. The SPICE rules reward plans that when they chop as little as possible into a county. This proves to be very valuable in urban areas that have dense populations: the chop may look small in area, but has a large impact in population. A macrochop is a chop that includes more than 5% of a district's quota. When a county is macrochopped in SC I will treat the school districts in that county as the subunits, since there are no townships as there are in the northeastern US. When a macrochop occurs, the school districts in that county act is if they were the counties in terms of chops and erosity.

Question: do these SPICE rules penalize my chop of Edisto Beach from Colleton County which I did because it only has a road connection to Charleston County?

[muon2]

I'm having trouble scoring the jimrtex plans here. They both seem to violate the 0.5% maximum deviation rule. I know earlier in the thread there were versions of his that did not, but I'm not sure what then is plan A and what is plan B.

jimrtex A



jimrtex B

[muon2]

I'm also having trouble scoring cvparty A. It appears that the Beaufort county part of CD 1 is only contiguous to the rest of the district by water contiguity without a bridge or ferry. That violates one of the map requirements.

Criteria for maps
★ REQUIRED ★
1) All maps are to be made with DRA
2) PVI data must be provided, and you must have your drf file ready to verify
3) Districts must be contiguous (water contiguity is allowed, but the areas should be connected by something like a bridge or ferry)
4) Populations must be provided; maximum allowed deviation is 0.5%

You must be logged in to read this quote.

[muon2]

I'll try to get the SPICE scores for the finalists here:

Plan	-S-	-P-	-I-	-C-	-E-
cvparty A
cvparty B
Aussie A
Single A
Single B
MB	2	11	10	12	39
True A	3	13	3	16	39
muon2 A	2	11	11	5	34
jimrtex A
jimrtex B

nb. The SPICE rules reward plans that when they chop as little as possible into a county. This proves to be very valuable in urban areas that have dense populations: the chop may look small in area, but has a large impact in population. A macrochop is a chop that includes more than 5% of a district's quota. When a county is macrochopped in SC I will treat the school districts in that county as the subunits, since there are no townships as there are in the northeastern US. When a macrochop occurs, the school districts in that county act is if they were the counties in terms of chops and erosity.

Question: do these SPICE rules penalize my chop of Edisto Beach from Colleton County which I did because it only has a road connection to Charleston County?

Whole contiguous counties are assumed to be self-connected even when they are not. The chop of Edisto beach does count as a county chop. Normally when a chop occurs there is a point of erosity for the connection between the two parts of the chop. However, since there is no road connecting across the chop boundary to Edisto there is no erosity there. So the answer is yes on the C score but no on the E score.

[cvparty]

I'm also having trouble scoring cvparty A. It appears that the Beaufort county part of CD 1 is only contiguous to the rest of the district by water contiguity without a bridge or ferry. That violates one of the map requirements.

Well, the literal requirement is that they just be connected by water. I added that it should have something like a bridge or ferry to guard against some wacky gerrymandered districts that are technically contiguous by water. I thought I connected Beaufort and Charleston by US 17 hmm, I'll update my map and add those two voting districts in Beaufort tho

[cvparty]

I'm having trouble scoring the jimrtex plans here. They both seem to violate the 0.5% maximum deviation rule. I know earlier in the thread there were versions of his that did not, but I'm not sure what then is plan A and what is plan B.

I believe his first two maps were his submissions

[muon2]

I'm having trouble scoring the jimrtex plans here. They both seem to violate the 0.5% maximum deviation rule. I know earlier in the thread there were versions of his that did not, but I'm not sure what then is plan A and what is plan B.

I believe his first two maps were his submissions

I was responding to the plans he copied on the post that had all ten. With changes along the way and plans spread over many pages it is extremely helpful to have a list of links or maps in a single post for reference.

[catographer]

My Georgia plan:

[jimrtex]

I'm having trouble scoring the jimrtex plans here. They both seem to violate the 0.5% maximum deviation rule. I know earlier in the thread there were versions of his that did not, but I'm not sure what then is plan A and what is plan B.

jimrtex A



jimrtex B

This variant of jimrtex A shifts 1 precinct in Darlington, 1 precinct in Lexington, and three precincts in Berkeley. I think it also shifts one precinct in Greenville. Standard deviation is 0.07%.

The variant of jimrtex B shifts the same five precincts, and one in Greenwood. It also has 0.07% standard deviation. I did not include a map.

The SCOTUS has said that there can be no de minimis deviation limit. Setting a 0.5% threshold is patently unconstitutional.

You can either accept the plan with the minimum number of county chops that provides a good faith attempt at practicable equality, or you are going to end up forcing to precise equality. In the latter case, you could use the number of persons victimized by county splits.

[muon2]

jimrtex raises a point that becomes even more important in GA. This exercise imposes a rule of a maximum deviation of 0.5% that may or not be what the court would impose, but it's a rule here. In some ways it's no different than saying that the VRA doesn't strictly apply. Panelists may consider it or not as they wish, even though the courts would absolutely require VRA compliance. In population SCOTUS often considers the range, not the maximum deviation (see Tennant v Jefferson county for a recent example).

In GA, Dekalb and Cobb counties are close to the quota for 1 CD. DeKalb county is no problem, but Cobb is 0.563% under the quota. A 0.5% maximum deviation guarantees a 1% range, but if no GA district exceeds 0.437% of the quota then a 1% range would still be maintained. A better rule might be a 1% range than a 0.5% maximum deviation. The alternative will be plans that tack one precinct onto Cobb just to bring its district up to quota.

The question for the panel is: Are plans that keep the population range within 1% acceptable, especially if they are used to avoid a chop?

[President Punxsutawney Phil]

I am inclined to say yes to that question.

[jimrtex]

jimrtex raises a point that becomes even more important in GA. This exercise imposes a rule of a maximum deviation of 0.5% that may or not be what the court would impose, but it's a rule here. In some ways it's no different than saying that the VRA doesn't strictly apply. Panelists may consider it or not as they wish, even though the courts would absolutely require VRA compliance. In population SCOTUS often considers the range, not the maximum deviation (see Tennant v Jefferson county for a recent example).

In GA, Dekalb and Cobb counties are close to the quota for 1 CD. DeKalb county is no problem, but Cobb is 0.563% under the quota. A 0.5% maximum deviation guarantees a 1% range, but if no GA district exceeds 0.437% of the quota then a 1% range would still be maintained. A better rule might be a 1% range than a 0.5% maximum deviation. The alternative will be plans that tack one precinct onto Cobb just to bring its district up to quota.

The question for the panel is: Are plans that keep the population range within 1% acceptable, especially if they are used to avoid a chop?

In that case use the simple jimrtex B which has a standard deviation of 0.37% and only splits Greenwood. For jimrtex A, use the version with Darlington, Lexington, and Dorchester splits, and a standard deviation of 0.07%.

If we are splitting counties, shouldn't we able to assume a perfect split, with the whole precinct version merely indicative?

[muon2]

jimrtex raises a point that becomes even more important in GA. This exercise imposes a rule of a maximum deviation of 0.5% that may or not be what the court would impose, but it's a rule here. In some ways it's no different than saying that the VRA doesn't strictly apply. Panelists may consider it or not as they wish, even though the courts would absolutely require VRA compliance. In population SCOTUS often considers the range, not the maximum deviation (see Tennant v Jefferson county for a recent example).

In GA, Dekalb and Cobb counties are close to the quota for 1 CD. DeKalb county is no problem, but Cobb is 0.563% under the quota. A 0.5% maximum deviation guarantees a 1% range, but if no GA district exceeds 0.437% of the quota then a 1% range would still be maintained. A better rule might be a 1% range than a 0.5% maximum deviation. The alternative will be plans that tack one precinct onto Cobb just to bring its district up to quota.

The question for the panel is: Are plans that keep the population range within 1% acceptable, especially if they are used to avoid a chop?

In that case use the simple jimrtex B which has a standard deviation of 0.37% and only splits Greenwood. For jimrtex A, use the version with Darlington, Lexington, and Dorchester splits, and a standard deviation of 0.07%.

If we are splitting counties, shouldn't we able to assume a perfect split, with the whole precinct version merely indicative?

Since we don't have political data I wouldn't assume a perfect split, though I suspect that would rarely if ever have a material effect on the partisan measures. It would matter if the chop is a macrochop and the county subunits come into play. In that case I could assume that the chops follow the subunit boundaries.

For jimrtex B I am reading the chops in a Anderson and Greenwood such that they keep the range within 1%.

[muon2]

I've completed scoring for the 10 SC plans. I've lined out plans that would be eliminated in all SPICE scenarios. If just geography (C, E) are considered then muon2 A and jimrtex B are the only two plans that survive elimination and are highlighted in green (jimrtex B is using the 1% range rule in lieu of the 0.5% max deviation rule). Since population is used as a final tie breaker, it shouldn't be used to eliminate plans, so INEQUALITY should not be considered in an elimination round, and plans that only survive by double counting population are in red. cvparty B would survive only if the political scores (S, P) are also used since it provide less skew.

I'll try to get the SPICE scores for the finalists here:

Plan	-S-	-P-	-I-	-C-	-E-
cvparty A	1	11	5	15	38
cvparty B	1	11	5	13	36
Aussie A	3	13	4	12	42
Single A	3	13	5	15	45
Single B	3	13	5	13	35
MB	2	11	10	12	39
True A	3	13	3	16	39
muon2 A	2	11	11	5	34
jimrtex A	4	10	5	7	39
jimrtex B	4	10	12	3	38

nb. The SPICE rules reward plans that when they chop as little as possible into a county. This proves to be very valuable in urban areas that have dense populations: the chop may look small in area, but has a large impact in population. A macrochop is a chop that includes more than 5% of a district's quota. When a county is macrochopped in SC I will treat the school districts in that county as the subunits, since there are no townships as there are in the northeastern US. When a macrochop occurs, the school districts in that county act is if they were the counties in terms of chops and erosity.

nb2. When a small chop goes into a county it is best to place it along the road that connects the county seats, or at least along a state or federal highway. The idea is that the fragments of a chopped county should be well connected to the other counties in the same district.

[jimrtex]

jimrtex raises a point that becomes even more important in GA. This exercise imposes a rule of a maximum deviation of 0.5% that may or not be what the court would impose, but it's a rule here. In some ways it's no different than saying that the VRA doesn't strictly apply. Panelists may consider it or not as they wish, even though the courts would absolutely require VRA compliance. In population SCOTUS often considers the range, not the maximum deviation (see Tennant v Jefferson county for a recent example).

In GA, Dekalb and Cobb counties are close to the quota for 1 CD. DeKalb county is no problem, but Cobb is 0.563% under the quota. A 0.5% maximum deviation guarantees a 1% range, but if no GA district exceeds 0.437% of the quota then a 1% range would still be maintained. A better rule might be a 1% range than a 0.5% maximum deviation. The alternative will be plans that tack one precinct onto Cobb just to bring its district up to quota.

The question for the panel is: Are plans that keep the population range within 1% acceptable, especially if they are used to avoid a chop?

In that case use the simple jimrtex B which has a standard deviation of 0.37% and only splits Greenwood. For jimrtex A, use the version with Darlington, Lexington, and Dorchester splits, and a standard deviation of 0.07%.

If we are splitting counties, shouldn't we able to assume a perfect split, with the whole precinct version merely indicative?

Since we don't have political data I wouldn't assume a perfect split, though I suspect that would rarely if ever have a material effect on the partisan measures. It would matter if the chop is a macrochop and the county subunits come into play. In that case I could assume that the chops follow the subunit boundaries.

For jimrtex B I am reading the chops in a Anderson and Greenwood such that they keep the range within 1%.

Oops, yes I forgot about Anderson.

My final little tweaks of one precinct in Darlington, one in Lexington, three in Dorchester moved the PVI about 0.1%. The population total of the three was around 7,000 so they aren't Macrochops. I calculated that I would have to move about 2500 persons across the state to make the plan perfect, this would have required an additional county split, since I would need six.

The problem that I see it, is that in moving the typical (one standard) deviation district from a range of 0.9943 to 1.0057, to 0.9993 to 1.0007 it will be seen as being eight times as good, rather 1/2 of 1% better. If someone maintained their weight within 1/2% of some target, they likely have an eating disorder.

[muon2]

The problem that I see it, is that in moving the typical (one standard) deviation district from a range of 0.9943 to 1.0057, to 0.9993 to 1.0007 it will be seen as being eight times as good, rather 1/2 of 1% better. If someone maintained their weight within 1/2% of some target, they likely have an eating disorder.

Oh oh. My weight is actually very stable and usually sits within a pound of a slowly moving average. A pound is only slightly more than 0.5% of my weight. Should I see someone?

[Dr. MB]

GA 1: R+16
GA 2: EVEN
GA 3: R+22
GA 4: D+25
GA 5: D+33
GA 6: R+16
GA 7: R+3
GA 8: R+10
GA 9: R+32
GA 10: R+13
GA 11: D+3
GA 12: D+1
GA 13: D+7
GA 14: R+29

Total: 7 R districts, 3 D districts, 4 swing districts

[Strudelcutie4427]

 

1. R+8 (-2432)
2. R+8 (-624)
3. R+17 (2032)
4. D+16 (473) 33% W, 34% B, 21% H
5. D+16 (-876) 34% W, 49% B, 12% H
6. D+30 (1199) 34% W, 55% B
7. R+14 (253)
8. R+10 (-1100)
9. R+30 (2349)
10. R+9 (55)
11. R+21 (-278)
12. R+10 (-1222)
13. D+15 (-1152) 34% W, 52% B
14. R+23 (1326)